CODEWITHSUNDEEP
pythonpandasscikit-learnclassificationone-hot-encoding
Bert Carremans
Bert Carremans

Reputation: 1733

How to use the output from OneHotEncoder in sklearn?

I have a Pandas Dataframe with 2 categorical variables, and ID variable and a target variable (for classification). I managed to convert the categorical values with OneHotEncoder. This results in a sparse matrix. 

ohe = OneHotEncoder()
# First I remapped the string values in the categorical variables to integers as OneHotEncoder needs integers as input
... remapping code ...

ohe.fit(df[['col_a', 'col_b']])
ohe.transform(df[['col_a', 'col_b']])

But I have no clue how I can use this sparse matrix in a DecisionTreeClassifier? Especially when I want to add some other non-categorical variables in my dataframe later on. Thanks!

EDIT In reply to the comment of miraculixx: I also tried the DataFrameMapper in sklearn-pandas

mapper = DataFrameMapper([
    ('id_col', None),
    ('target_col', None),
    (['col_a'], OneHotEncoder()),
    (['col_b'], OneHotEncoder())
])

t = mapper.fit_transform(df)

But then I get this error:

TypeError: no supported conversion for types : (dtype('O'), dtype('int64'), dtype('float64'), dtype('float64')).

Upvotes: 8

Views: 10561

Answers (2)

Merlin
Merlin

Reputation: 25639

Look at this example from scikit-learn: http://scikit-learn.org/stable/auto_examples/ensemble/plot_feature_transformation.html#example-ensemble-plot-feature-transformation-py

Problem is that you are not using the sparse matrices to xx.fit(). You are using the original data.

Upvotes: 1

Guiem Bosch
Guiem Bosch

Reputation: 2758

I see you are already using Pandas, so why not using its get_dummies function?

import pandas as pd
df = pd.DataFrame([['rick','young'],['phil','old'],['john','teenager']],columns=['name','age-group'])

result

   name age-group
0  rick     young
1  phil       old
2  john  teenager

now you encode with get_dummies

pd.get_dummies(df)

result

name_john  name_phil  name_rick  age-group_old  age-group_teenager  \
0          0          0          1              0                   0   
1          0          1          0              1                   0   
2          1          0          0              0                   1   

   age-group_young  
0                1  
1                0  
2                0

And you can actually use the new Pandas DataFrame in your Sklearn's DecisionTreeClassifier.

Upvotes: 14

Related Questions